Apple has repeatedly supported user privacy through past technical implementations and policy initiatives. However, Apple took a large and very public step in the opposite direction on Aug. 5. The company provided a technical solution that it believes would simultaneously continue to protect the privacy of peoples’ communications and devices while allowing the company to track those who share child sexual abuse material (CSAM). But in a conflict of two social goods—providing privacy and security of communications and stored data, and preventing harm to children—Apple has made a gamble that normalizes surveillance.
…Apple’s new definition of end-to-end encryption (E2E) means that Apple tools could have access to any decrypted contents. Read More
Monthly Archives: August 2021
Machine learning’s crumbling foundations
Technological debt is insidious, a kind of socio-infrastructural subprime crisis that’s unfolding around us in slow motion. Our digital infrastructure is built atop layers and layers and layers of code that’s insecure due to a combination of bad practices and bad frameworks.
Even people who write secure code import insecure libraries, or plug it into insecure authorization systems or databases. Like asbestos in the walls, this cruft has been fragmenting, drifting into our air a crumb at a time.
We ignored these, treating them as containable, little breaches and now the walls are rupturing and choking clouds of toxic waste are everywhere. Read More
Headline or Trend Line? Evaluating Chinese-Russian Collaboration in AI
China and Russia have declared 2020 and 2021 as years of scientific and technological innovation cooperation, focusing on biotech, artificial intelligence, and robotics.1 Both countries view AI as critical to their respective domestic and foreign policy objectives, and they are ramping up investments in AI-related research and development, though China’s investments far outweigh Russia’s. U.S. observers are watching this convergence between America’s two key competitors with increasing concern, if not alarm. Some worry that alignment between Beijing and Moscow, especially in the areas of science and technology, could accelerate the development of surveillance tools to enhance authoritarian control of domestic populations. Others warn that deepening Sino-Russian cooperation will dilute the effects of sanctions on Russia. Still others fear that the strengthening partnership between China and Russia will undermine U.S. strategic interests and those of its democratic allies in Europe and Asia.2
Chinese and Russian sources are keen to publicize their “comprehensive strategic partnership of coordination for the new era,” potentially underscoring the seriousness of their joint ambitions.3 Yet the scale and scope of this emerging partnership deserve closer scrutiny, particularly in the field of AI. To what extent are China and Russia following up on their declared intentions to foster joint research, development, and commercialization of AI-related technologies? In other words, how do we separate headlines from trend lines?
This issue brief analyzes the scope of cooperation and relative trends between China and Russia in two key metrics of AI development: research publications and investment. Read More
GAN Objective Functions: GANs and Their Variations
There are hundreds of types of GANs. How does an objective function play into what a GAN looks like?
If you haven’t already, you should definitely read my previous post about what a GAN is (especially if you don’t know what I mean when I say GAN!). That post should give you a starting point to dive into the world of GANs and how they work. It’s a solid primer for any article on GANs, not to mention this one where we will be discussing objective functions of GANs and other variations of GANs currently out there that use twists on defining their objectives for different results. Read More
The Future of AI in 2025 and Beyond
By 2025, artificial intelligence (AI) will significantly improve our daily life by handling some of today’s complex tasks with great efficiency.
The leading AI researcher, Geoff Hinton, stated that it is very hard to predict what advances AI will bring beyond five years, noting that exponential progress makes the uncertainty too great.
This article will therefore consider both the opportunities as well as the challenges that we will face along the way across different sectors of the economy. It is not intended to be exhaustive. Read More
Spies for Hire: China’s New Breed of Hackers Blends Espionage and Entrepreneurship
The state security ministry is recruiting from a vast pool of private-sector hackers who often have their own agendas and sometimes use their access for commercial cybercrime, experts say.
China’s buzzy high-tech companies don’t usually recruit Cambodian speakers, so the job ads for three well-paid positions with those language skills stood out. The ad, seeking writers of research reports, was placed by an internet security start-up in China’s tropical island-province of Hainan.
That start-up was more than it seemed, according to American law enforcement. Hainan Xiandun Technology was part of a web of front companies controlled by China’s secretive state security ministry, according to a federal indictment from May. They hacked computers from the United States to Cambodia to Saudi Arabia, seeking sensitive government data as well as less-obvious spy stuff, like details of a New Jersey company’s fire-suppression system, according to prosecutors. Read More
Toward a rational and ethical sociotechnical system of autonomous vehicles: A novel application of multi-criteria decision analysis
The impacts of autonomous vehicles (AV) are widely anticipated to be socially, economically, and ethically significant. A reliable assessment of the harms and benefits of their large-scale deployment requires a multi-disciplinary approach. To that end, we employed Multi-Criteria Decision Analysis to make such an assessment. We obtained opinions from 19 disciplinary experts to assess the significance of 13 potential harms and eight potential benefits that might arise under four deployments schemes. Specifically, we considered: (1) the status quo, i.e., no AVs are deployed; (2) unfettered assimilation, i.e., no regulatory control would be exercised and commercial entities would “push” the development and deployment; (3) regulated introduction, i.e., regulatory control would be applied and either private individuals or commercial fleet operators could own the AVs; and (4) fleets only, i.e., regulatory control would be applied and only commercial fleet operators could own the AVs. Our results suggest that two of these scenarios, (3) and (4), namely regulated privately-owned introduction or fleet ownership or autonomous vehicles would be less likely to cause harm than either the status quo or the unfettered options. Read More
How Data Brokers Sell Access to the Backbone of the Internet
ISPs are quietly distributing “netflow” data that can, among other things, trace traffic through VPNs.
There’s something of an open secret in the cybersecurity world: internet service providers quietly give away detailed information about which computer is communicating with another to private businesses, which then sells access to that data to a range of third parties, according to multiple sources in the threat intelligence industry.
The information, known as netflow data, is a useful tool for digital investigators. They can use it to identify servers being used by hackers, or to follow data as it is stolen. But the sale of this information still makes some people nervous because they are concerned about whose hands it may fall into. Read More
VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning
Recent self-supervised methods for image representation learning are based on maximizing the agreement between embedding vectors from different views of the same image. A trivial solution is obtained when the encoder outputs constant vectors. This collapse problem is often avoided through implicit biases in the learning architecture, that often lack a clear justification or interpretation. In this paper, we introduce VICReg (Variance-Invariance-Covariance Regularization), a method that explicitly avoids the collapse problem with a simple regularization term on the variance of the embeddings along each dimension individually. VICReg combines the variance term with a decorrelation mechanism based on redundancy reduction and covariance regularization, and achieves results on par with the state of the art on several downstream tasks. In addition, we show that incorporating our new variance term into other methods helps stabilize the training and leads to performance improvements. Read More
#image-recognition, #self-supervisedUnity acquires AI chat analysis platform Oto, launches toxicity in gaming report
Unity has acquired Oto, an AI-based audio chat analysis platform that figures out if humans need to intercede in a toxic multiplayer gaming environment. The companies did not disclose terms.
We all know that online gamers can be toxic, trash-talking each other. Some of that is OK and can be chalked up to the culture around a game. But some of it also crosses the line, and that’s where Oto comes in. Read More
Rick's Cafe AI 